---
title: "Trace2Skill — paper by Ni et al. (2026) — summary and ideas"
description: "LLM agents increasingly depend on skill files, which are reusable documents… 57.65 points on WikiTQ. 42 references, cited by 11 papers here."
authors: ["Jingwei Ni", "Yihao Liu", "Xinpeng Liu", "Yutao Sun", "Mengyu Zhou", "Pengyu Cheng", "Dexin Wang", "Erchao Zhao", "Xiaoxi Jiang", "Guanjun Jiang"]
year: 2026
venue: "arXiv:2603.25158"
source: "https://arxiv.org/abs/2603.25158"
doi: "10.48550/arxiv.2603.25158"
arxiv: "2603.25158"
licence: "CC BY 4.0 (arXiv)"
licence_url: "http://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/ni-2026-trace2skill-distill-trajectory-local-lessons/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/ni-2026-trace2skill-distill-trajectory-local-lessons/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/) / Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills

# Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills

**Authors:** [Jingwei Ni](https://md.research.elamur.ai/free/en/authors/jingwei-ni/), [Yihao Liu](https://md.research.elamur.ai/free/en/authors/yihao-liu/), Xinpeng Liu, [Yutao Sun](https://md.research.elamur.ai/free/en/authors/yutao-sun/), [Mengyu Zhou](https://md.research.elamur.ai/free/en/authors/mengyu-zhou/), Pengyu Cheng, Dexin Wang, Erchao Zhao, [Xiaoxi Jiang](https://md.research.elamur.ai/free/en/authors/xiaoxi-jiang/) and Guanjun Jiang · **Year:** 2026 · **Venue:** arXiv:2603.25158 · **Licence:** [CC BY 4.0 (arXiv)](http://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2603.25158) · [PDF](https://arxiv.org/pdf/2603.25158) · [DOI 10.48550/arxiv.2603.25158](https://doi.org/10.48550/arxiv.2603.25158) · [arXiv 2603.25158](https://arxiv.org/abs/2603.25158) · [OpenAlex](https://openalex.org/W7141388408)

**Citations:** 1 citations (OpenAlex, 2026-10-14) · [11](https://md.research.elamur.ai/free/en/papers/ni-2026-trace2skill-distill-trajectory-local-lessons/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
LLM agents increasingly depend on skill files, which are reusable documents describing procedures for a domain. Writing these by hand is costly. Drafts produced from a model's own knowledge tend to omit the practical pitfalls that only show up during real execution. Trace2Skill learns skills from agent experience in three stages:

1. A frozen agent runs ReAct-style on an evolving task set, and its trajectories are labeled as successes or failures.
2. Separate analyst sub-agents examine each trajectory in parallel and propose a patch. Failures go to an agentic analyst that can inspect artifacts and test fixes.
3. All patches are merged hierarchically into a single skill directory. Lessons that recur across patches are kept, and one-off edits are dropped.

The same model plays every role, so no teacher model is needed. The framework can refine an expert-written skill or build one from a weak draft. On SpreadsheetBench-Verified, the evolved skills improve held-out performance. They also carry over to other model sizes, other model families, out-of-distribution table QA, math reasoning, DocVQA, and office-document tasks. The result is a static skill that requires neither retraining nor test-time retrieval.

## Key points
- **Spreadsheet setup:** SpreadsheetBench-Verified (400 tasks) is split into 200 evolving and 200 test problems. WikiTableQuestions and HiTab serve as OOD transfer targets.
- **Cross-model transfer:** Skills evolved from Qwen3.5-35B-A3B traces raised a Qwen3.5-122B-A10B agent by up to 57.65 points on WikiTQ. A skill evolved with Gemma-4-31B-it also helped Qwen3.5-122B.
- **Mismatch of the official skill:** Anthropic's xlsx skill lifted the 122B model's Vrf score from 27.67 to 48.33. The same skill lowered the 35B model's score from 19.00 to 9.67.
- **Other office domains (starting from Anthropic's official skills):**
  - PDF: 76.9% → 85.3% (VRDU traces, evaluated on VAREX)
  - PPTX: 72.5% → 88.8% (TSBench, evaluated on an OOD split)
  - DOCX: 79.7% → 87.5% (synthetic tasks, evaluated on OfficeBench)
- **Comparisons and limitations:**
  - Parallel consolidation is faster than sequential skill editing and usually stronger.
  - It also outperforms ReasoningBank-style retrieval memory.
  - Selecting patches with Bayesian optimization can help further but is expensive.
  - Code is available at github.com/Qwen-Applications/Trace2Skill.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [Peking University](https://md.research.elamur.ai/free/en/universities/peking-university/) — University, Beijing, China · 335 works · 43,758 citations
- [Alibaba Group](https://md.research.elamur.ai/free/en/companies/alibaba-group/) — Company, Hangzhou, China · 207 works · 17,449 citations
- [Zhejiang University](https://md.research.elamur.ai/free/en/universities/zhejiang-university/) — University, Hangzhou, China · 236 works · 14,072 citations
- [ETH Zurich](https://md.research.elamur.ai/free/en/universities/eth-zurich/) — University, Zurich, Switzerland · 97 works · 9,612 citations
- [University of Zurich](https://md.research.elamur.ai/free/en/universities/university-of-zurich/) — University, Zurich, Switzerland · 83 works · 12,449 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 19,503 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 18,805 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,959 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,789 works

## Cite

Cite as: Jingwei Ni, Yihao Liu, Xinpeng Liu, Yutao Sun, Mengyu Zhou, Pengyu Cheng, Dexin Wang, Erchao Zhao, Xiaoxi Jiang and Guanjun Jiang. 2026. Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills. arXiv:2603.25158.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@misc{ni2026trace2skill,
  title = {{Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills}},
  author = {Jingwei Ni, Yihao Liu, Xinpeng Liu, Yutao Sun, Mengyu Zhou, Pengyu Cheng, Dexin Wang, Erchao Zhao, Xiaoxi Jiang and Guanjun Jiang},
  year = {2026},
  doi = {10.48550/arxiv.2603.25158},
  eprint = {2603.25158},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2603.25158}
}
```

- [← previous Spatial-RAG: Spatial Retrieval Augmented Generation for Real-World Geospatial Reasoning Questions](https://md.research.elamur.ai/free/en/papers/yu-2026-spatial-rag-spatial-retrieval-augmented/)
- [next → A Principled Framework for Evaluating on Typologically Diverse Languages](https://md.research.elamur.ai/free/en/papers/ploeger-2026-principled-framework-evaluating-typologically-diverse/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/)

## References

42 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Salaheddin Alzubi, Noah Provenzano, Jaydon Bingham, Weiyuan Chen, and Tu Vu. Evoskill: Automated skill discovery for multi-agent systems, 2026. URL https://arxiv.org/abs/2603.02766. — [arXiv:2603.02766](https://arxiv.org/abs/2603.02766) [↗](https://arxiv.org/abs/2603.02766)
2. Anthropic. PDF, DOCX, and PPTX document skills. GitHub repository, 2026a. URL [https://github.com/](https://github.com/) anthropics/skills/tree/main/skills. Accessed: 2026-05-01; document-skill subfolders: pdf, docx, and pptx. [↗](https://github.com/)
3. Anthropic. How to create a skill with claude through conversation. Claude Tutorials, 2026b. URL [https://claude.com/resources/tutorials/](https://claude.com/resources/tutorials/) how-to-create-a-skill-with-claude-through-conversation. [↗](https://claude.com/resources/tutorials/)
4. Anthropic. skill-creator. GitHub repository, 2026c. URL [https://github.com/anthropics/skills/](https://github.com/anthropics/skills/) tree/main/skills/skill-creator. [↗](https://github.com/anthropics/skills/)
5. Anthropic. What are skills? Claude Help Center, 2026d. URL [https://support.claude.com/en/](https://support.claude.com/en/) articles/12512176-what-are-skills. Access Date: 2026-03-22. [↗](https://support.claude.com/en/)
6. Udi Barzelay, Ophir Azulai, Inbar Shapira, Idan Friedman, Foad Abo Dahood, Madison Lee, and Abraham Daniels. VAREX: A benchmark for multi-modal structured extraction from documents, 2026. URL https://arxiv.org/abs/2603.15118. — [arXiv:2603.15118](https://arxiv.org/abs/2603.15118) [↗](https://arxiv.org/abs/2603.15118)
7. Zhoujun Cheng, Haoyu Dong, Zhiruo Wang, Ran Jia, Jiaqi Guo, Yan Gao, Shi Han, Jian-Guang Lou, and Dongmei Zhang. HiTab: A hierarchical table dataset for question answering and natural language generation. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Vol… [find](https://scholar.google.com/scholar?q=Zhoujun+Cheng%2C+Haoyu+Dong%2C+Zhiruo+Wang%2C+Ran+Jia%2C+Jiaqi+Guo%2C+Yan+Gao%2C+Shi+Han%2C+Jian%2DGuang+Lou%2C+and+Dongmei+Zhang%2E+HiTab%3A+A+hierarchical+table+dataset+for+question+answering+and+natural+language+generat)
8. Runnan Fang, Yuan Liang, Xiaobin Wang, Jialong Wu, Shuofei Qiao, Pengjun Xie, Fei Huang, Huajun Chen, and Ningyu Zhang. Memp: Exploring agent procedural memory, 2026. URL https://arxiv.org/abs/ 2508.06433. [↗](https://arxiv.org/abs/)
9. Google. Gemma 4: Byte for byte, the most capable open models. [https://blog.google/](https://blog.google/) innovation-and-ai/technology/developers-tools/gemma-4/, 2026. Accessed: 2026-05-18. [↗](https://blog.google/)
10. Tingxu Han, Yi Zhang, Wei Song, Chunrong Fang, Zhenyu Chen, Youcheng Sun, and Lijie Hu. Swe-skills-bench: Do agent skills actually help in real-world software engineering?, 2026. URL https://arxiv.org/abs/ 2603.15401. [↗](https://arxiv.org/abs/)
11. Guanyu Jiang, Zhaochen Su, Xiaoye Qu, and Yi R. Fung. Xskill: Continual learning from experience and skills in multimodal agents, 2026. URL https://arxiv.org/abs/2603.12056. — [arXiv:2603.12056](https://arxiv.org/abs/2603.12056) [↗](https://arxiv.org/abs/2603.12056)
12. Kyudan Jung, Hojun Cho, Jooyeol Yun, Soyoung Yang, Jaehyeok Jang, and Jaegul Choo. Talk to your slides: High-efficiency slide editing via language-driven structured data manipulation, 2026. URL https://arxiv.org/abs/2505.11604. [↗](https://arxiv.org/abs/2505.11604)
13. Woosuk Kwon, Zhuohan Li, Siyuan Zhuang, Ying Sheng, Lianmin Zheng, Cody Hao Yu, Joseph E. Gonzalez, Hao Zhang, and Ion Stoica. [Efficient memory management for large language model serving with pagedattention](https://md.research.elamur.ai/free/en/papers/kwon-2023-efficient-memory-management-large-language/), 2023. URL https://arxiv.org/abs/2309.06180. [↗](https://arxiv.org/abs/2309.06180)
14. Hao Li, Chunjiang Mu, Jianhao Chen, Siyue Ren, Zhiyao Cui, Yiqun Zhang, Lei Bai, and Shuyue Hu. Organizing, orchestrating, and benchmarking agent skills at ecosystem scale, 2026a. URL https://arxiv.org/abs/ 2603.02176. [↗](https://arxiv.org/abs/)
15. Jiachun Li, Pengfei Cao, Zhuoran Jin, Yubo Chen, Kang Liu, and Jun Zhao. Mirage: Evaluating and explaining inductive reasoning process in language models, 2025. URL https://arxiv.org/abs/2410.09542. — [arXiv:2410.09542](https://arxiv.org/abs/2410.09542) [↗](https://arxiv.org/abs/2410.09542)
16. Xiangyi Li, Wenbo Chen, Yimin Liu, Shenghan Zheng, Xiaokun Chen, Yifeng He, Yubo Li, Bingran You, Haotian Shen, Jiankai Sun, Shuyi Wang, Binxu Li, Qunhong Zeng, Di Wang, Xuandong Zhao, Yuanli Wang, Roey Ben Chaim, Zonglin Di, Yipeng Gao, Junwei He, Yizhuo He, Liqiang Jing, Luyang Kong, Xin Lan, Jiac… — [arXiv:2602.12670](https://arxiv.org/abs/2602.12670) [find](https://scholar.google.com/scholar?q=Xiangyi+Li%2C+Wenbo+Chen%2C+Yimin+Liu%2C+Shenghan+Zheng%2C+Xiaokun+Chen%2C+Yifeng+He%2C+Yubo+Li%2C+Bingran+You%2C+Haotian+Shen%2C+Jiankai+Sun%2C+Shuyi+Wang%2C+Binxu+Li%2C+Qunhong+Zeng%2C+Di+Wang%2C+Xuandong+Zhao%2C+Yuanli+Wang%2C+Ro)
17. Xiaoxiao Li. When single-agent with skills replace multi-agent systems and when they fail, 2026. URL https://arxiv.org/abs/2601.04748. — [arXiv:2601.04748](https://arxiv.org/abs/2601.04748) [↗](https://arxiv.org/abs/2601.04748)
18. Yu Li, Rui Miao, Zhengling Qi, and Tian Lan. Arise: Agent reasoning with intrinsic skill evolution in hierarchical reinforcement learning, 2026c. URL https://arxiv.org/abs/2603.16060. — [arXiv:2603.16060](https://arxiv.org/abs/2603.16060) [↗](https://arxiv.org/abs/2603.16060)
19. Yuan Liang, Ruobin Zhong, Haoming Xu, Chen Jiang, Yi Zhong, Runnan Fang, Jia-Chen Gu, Shumin Deng, Yunzhi Yao, Mengru Wang, Shuofei Qiao, Xin Xu, Tongtong Wu, Kun Wang, Yang Liu, Zhen Bi, Jungang Lou, Yuchen Eleanor Jiang, Hangcheng Zhu, Gang Yu, Haiwen Hong, Longtao Huang, Hui Xue, Chenxi Wang, Yij… — [arXiv:2603.04448](https://arxiv.org/abs/2603.04448) [find](https://scholar.google.com/scholar?q=Yuan+Liang%2C+Ruobin+Zhong%2C+Haoming+Xu%2C+Chen+Jiang%2C+Yi+Zhong%2C+Runnan+Fang%2C+Jia%2DChen+Gu%2C+Shumin+Deng%2C+Yunzhi+Yao%2C+Mengru+Wang%2C+Shuofei+Qiao%2C+Xin+Xu%2C+Tongtong+Wu%2C+Kun+Wang%2C+Yang+Liu%2C+Zhen+Bi%2C+Jungang+Lou%2C)
20. Brian S. Lin, Jiaxin Yuan, Zihan Zhou, Shouli Wang, Shuo Wang, Cunliang Kong, Qi Shi, Yuxuan Li, Liner Yang, Zhiyuan Liu, and Maosong Sun. On llm-based scientific inductive reasoning beyond equations, 2025. URL https://arxiv.org/abs/2509.16226. [↗](https://arxiv.org/abs/2509.16226)
21. Yitao Liu, Chenglei Si, Karthik Narasimhan, and Shunyu Yao. Contextual experience replay for self-improvement of language agents, 2025. URL https://arxiv.org/abs/2506.06698. [↗](https://arxiv.org/abs/2506.06698)
22. Yuchen Ma, Yue Huang, Han Bao, Haomin Zhuang, Swadheen Shukla, Michel Galley, Xiangliang Zhang, and Stefan Feuerriegel. Skillgen: Verified inference-time agent skill synthesis, 2026. URL https://arxiv.org/abs/2605.10999. — [arXiv:2605.10999](https://arxiv.org/abs/2605.10999) [↗](https://arxiv.org/abs/2605.10999)
23. Zeyao Ma, Bohan Zhang, Jing Zhang, Jifan Yu, Xiaokang Zhang, Xiaohan Zhang, Sijia Luo, Xi Wang, and Jie Tang. [Spreadsheetbench: Towards challenging real world spreadsheet manipulation](https://md.research.elamur.ai/free/en/papers/ma-2024-spreadsheetbench-towards-challenging-real-world/), 2024. URL https://arxiv.org/abs/2406.14991. [↗](https://arxiv.org/abs/2406.14991)
24. Minesh Mathew, Dimosthenis Karatzas, R Manmatha, and CV Jawahar. [Docvqa: A dataset for vqa on document images](https://md.research.elamur.ai/free/en/papers/mathew-2020-docvqa-dataset-vqa-document-images/). corr abs/2007.00398 (2020). arXiv preprint arXiv:2007.00398, 2020.
25. Kolby Nottingham, Bodhisattwa Prasad Majumder, Bhavana Dalvi Mishra, Sameer Singh, Peter Clark, and Roy Fox. Skill set optimization: Reinforcing language model behavior via transferable skills, 2024. URL https://arxiv.org/abs/2402.03244. — [arXiv:2402.03244](https://arxiv.org/abs/2402.03244) [↗](https://arxiv.org/abs/2402.03244)
26. Siru Ouyang, Jun Yan, I-Hung Hsu, Yanfei Chen, Ke Jiang, Zifeng Wang, Rujun Han, Long Le, Samira Daruki, Xiangru Tang, Vishy Tirumalashetty, George Lee, Mahsan Rofouei, Hangfei Lin, Jiawei Han, Chen-Yu Lee, and Tomas Pfister. Reasoningbank: Scaling agent self-evolving with reasoning memory. In The F… [find](https://scholar.google.com/scholar?q=Siru+Ouyang%2C+Jun+Yan%2C+I%2DHung+Hsu%2C+Yanfei+Chen%2C+Ke+Jiang%2C+Zifeng+Wang%2C+Rujun+Han%2C+Long+Le%2C+Samira+Daruki%2C+Xiangru+Tang%2C+Vishy+Tirumalashetty%2C+George+Lee%2C+Mahsan+Rofouei%2C+Hangfei+Lin%2C+Jiawei+Han%2C+Chen%2DY)
27. Panupong Pasupat and Percy Liang. [Compositional semantic parsing on semi-structured tables](https://md.research.elamur.ai/free/en/papers/pasupat-2015-compositional-semantic-parsing-semi-structured/), 2015. URL https://arxiv.org/abs/1508.00305. [↗](https://arxiv.org/abs/1508.00305)
28. Cheng Qian, Shihao Liang, Yujia Qin, Yining Ye, Xin Cong, Yankai Lin, Yesai Wu, Zhiyuan Liu, and Maosong Sun. Investigate-consolidate-exploit: A general strategy for inter-task agent self-evolution, 2024. URL https://arxiv.org/abs/2401.13996. — [arXiv:2401.13996](https://arxiv.org/abs/2401.13996) [↗](https://arxiv.org/abs/2401.13996)
29. Noah Shinn, Federico Cassano, Edward Berman, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao. [Reflexion: Language agents with verbal reinforcement learning](https://md.research.elamur.ai/free/en/papers/shinn-2023-reflexion-language-agents-verbal-reinforcement/), 2023. URL https://arxiv.org/abs/ 2303.11366. [↗](https://arxiv.org/abs/)
30. Qwen Team. Qwen3.5: Accelerating productivity with native multimodal agents, February 2026. URL [https://qwen.ai/blog?id=qwen3.5](https://qwen.ai/blog?id=qwen3.5). [↗](https://qwen.ai/blog?id=qwen3.5)
31. Guanzhi Wang, Yuqi Xie, Yunfan Jiang, Ajay Mandlekar, Chaowei Xiao, Yuke Zhu, Linxi Fan, and Anima Anandkumar. Voyager: An open-ended embodied agent with large language models, 2023a. URL https://arxiv.org/abs/2305.16291. [↗](https://arxiv.org/abs/2305.16291)
32. Zilong Wang, Yichao Zhou, Wei Wei, Chen-Yu Lee, and Sandeep Tata. VRDU: A benchmark for visually-rich document understanding. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2023b. doi: 10.1145/3580305.3599929. URL https://arxiv.org/abs/2211.15421. — [doi:10.1145/3580305.3599929](https://doi.org/10.1145/3580305.3599929) [↗](https://arxiv.org/abs/2211.15421)
33. Zilong Wang, Yuedong Cui, Li Zhong, Zimin Zhang, Da Yin, Bill Yuchen Lin, and Jingbo Shang. OfficeBench: Benchmarking language agents across multiple applications for office automation, 2024a. URL https://arxiv.org/abs/2407.19056. — [arXiv:2407.19056](https://arxiv.org/abs/2407.19056) [↗](https://arxiv.org/abs/2407.19056)
34. Zora Zhiruo Wang, Jiayuan Mao, Daniel Fried, and Graham Neubig. [Agent workflow memory](https://md.research.elamur.ai/free/en/papers/wang-2024-agent-workflow-memory/), 2024b. URL https://arxiv.org/abs/2409.07429. [↗](https://arxiv.org/abs/2409.07429)
35. Peng Xia, Jianwen Chen, Hanyang Wang, Jiaqi Liu, Kaide Zeng, Yu Wang, Siwei Han, Yiyang Zhou, Xujiang Zhao, Haifeng Chen, Zeyu Zheng, Cihang Xie, and Huaxiu Yao. Skillrl: Evolving agents via recursive skill-augmented reinforcement learning, 2026a. URL https://arxiv.org/abs/2602.08234. — [arXiv:2602.08234](https://arxiv.org/abs/2602.08234) [↗](https://arxiv.org/abs/2602.08234)
36. Peng Xia, Jianwen Chen, Xinyu Yang, Haoqin Tu, Jiaqi Liu, Kaiwen Xiong, Siwei Han, Shi Qiu, Haonian Ji, Yuyin Zhou, Zeyu Zheng, Cihang Xie, and Huaxiu Yao. Metaclaw: Just talk – an agent that meta-learns and evolves in the wild, 2026b. URL https://arxiv.org/abs/2603.17187. [↗](https://arxiv.org/abs/2603.17187)
37. Chenfei Xiong, Jingwei Ni, Yu Fan, Vilém Zouhar, Donya Rooein, Lorena Calvo-Bartolomé, Alexander Hoyle, Zhijing Jin, Mrinmaya Sachan, Markus Leippold, Dirk Hovy, Mennatallah El-Assady, and Elliott Ash. Co-DETECT: Collaborative discovery of edge cases in text classification. In Ivan Habernal, Peter S… [find](https://scholar.google.com/scholar?q=Chenfei+Xiong%2C+Jingwei+Ni%2C+Yu+Fan%2C+Vil%C3%A9m+Zouhar%2C+Donya+Rooein%2C+Lorena+Calvo%2DBartolom%C3%A9%2C+Alexander+Hoyle%2C+Zhijing+Jin%2C+Mrinmaya+Sachan%2C+Markus+Leippold%2C+Dirk+Hovy%2C+Mennatallah+El%2DAssady%2C+and+Elliott+Ash)
38. Yutao Yang, Junsong Li, Qianjun Pan, Bihao Zhan, Yuxuan Cai, Lin Du, Jie Zhou, Kai Chen, Qin Chen, Xin Li, Bo Zhang, and Liang He. Autoskill: Experience-driven lifelong learning via skill self-evolution, 2026. URL https://arxiv.org/abs/2603.01145. — [arXiv:2603.01145](https://arxiv.org/abs/2603.01145) [↗](https://arxiv.org/abs/2603.01145)
39. Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. [React: Synergizing reasoning and acting in language models](https://md.research.elamur.ai/free/en/papers/yao-2022-react-synergizing-reasoning-acting-language/), 2023. URL https://arxiv.org/abs/2210. 03629. [↗](https://arxiv.org/abs/2210)
40. Qiying Yu, Zheng Zhang, Ruofei Zhu, Yufeng Yuan, Xiaochen Zuo, Yu Yue, Weinan Dai, Tiantian Fan, Gaohong Liu, Lingjun Liu, Xin Liu, Haibin Lin, Zhiqi Lin, Bole Ma, Guangming Sheng, Yuxuan Tong, Chi Zhang, Mofan Zhang, Wang Zhang, Hang Zhu, Jinhua Zhu, Jiaze Chen, Jiangjie Chen, Chengyi Wang, Hongli… — [DAPO: An Open-Source LLM Reinforcement Learning System at Scale](https://md.research.elamur.ai/free/en/papers/yu-2025-dapo-open-source-llm-reinforcement/)
41. Boyuan Zheng, Michael Y. Fatemi, Xiaolong Jin, Zora Zhiruo Wang, Apurva Gandhi, Yueqi Song, Yu Gu, Jayanth Srinivasa, Gaowen Liu, Graham Neubig, and Yu Su. Skillweaver: Web agents can self-improve by discovering and honing skills, 2025. URL https://arxiv.org/abs/2504.07079. [↗](https://arxiv.org/abs/2504.07079)
42. Huichi Zhou, Siyuan Guo, Anjie Liu, Zhongwei Yu, Ziqin Gong, Bowen Zhao, Zhixun Chen, Menglong Zhang, Yihang Chen, Jinsong Li, Runyu Yang, Qiangbin Liu, Xinlei Yu, Jianmin Zhou, Na Wang, Chunyang Sun, and Jun Wang. Memento-skills: Let agents design agents, 2026a. URL https://arxiv.org/abs/2603. 1874… [↗](https://arxiv.org/abs/2603)

*Updated 3 November 2026*